mirror of https://github.com/apache/lucene.git
doc: explain why peersync fingerprinting is sometimes off
This commit is contained in:
parent
091889cf79
commit
5d2fb9cf69
|
@ -173,6 +173,9 @@ public class SyncStrategy {
|
|||
// if we can't reach a replica for sync, we still consider the overall sync a success
|
||||
// TODO: as an assurance, we should still try and tell the sync nodes that we couldn't reach
|
||||
// to recover once more?
|
||||
// Fingerprinting here is off because the we currently rely on having at least one of the nodes return "true", and if replicas are out-of-sync
|
||||
// we still need to pick one as leader. A followup sync from the replica to the new leader (with fingerprinting on) should then fail and
|
||||
// initiate recovery-by-replication.
|
||||
PeerSync peerSync = new PeerSync(core, syncWith, core.getUpdateHandler().getUpdateLog().getNumRecordsToKeep(), true, true, peerSyncOnlyWithActive, false);
|
||||
return peerSync.sync();
|
||||
}
|
||||
|
|
Loading…
Reference in New Issue